Research On Digital Watermarking Collusion Attack Detection Algorithm Based On The Semi-Definite Relaxation

In the era of Internet, infringement and illegal copies threat the spread of multimedia products. Applying digital watermark in products can help to prevent illegal spread effectively. Watermark technique, which is a cross-field of image recognition, signal process, e-commerce, cryptography and other disciplines, is of significant research value. As one of the information hiding techniques, watermark adds specific identification and verification information by using redundant bits in medium data to protect copyright.Unauthorized users must erase or fudge the watermark to trade digital products freely. Thus, watermark detection and verification are very important parts in watermark system. It requires to detect illegal copies quickly and correctly under certain complexity and to offer necessary technical support for copyright prosecution. Nowadays, the most effective attack to watermark is collusion, which can erase or reduce fingerprint information by combing several legal users’copy files. Classical digital watermarks usually take anti-collusion performance into account when design the algorithms, such as C-security code, orthogonal watermark coding, and "logical AND by default" anti-collusion coding. WBE (Welch Bound Equality) Sequence has good performance and is widely used in signal coding. Considering that it can satisfy many requirements of watermark coding, we discuss to bring WBE coding idea into digital watermark design.In the view of signal processing, digital watermark insertion and detection in system are just like signal transmission and detection. Based on the collusion watermark signal at receiver, system can trace the collusion users. Actually, it is the process of finding the user state set which has the biggest relativity with collusion signal. The best retrieving method is most likelihood algorithm. Combine with signal processing knowledge, this thesis proposes a solution of using communication detection algorithm into digital watermark detection.In this thesis, digital watermark system is simulated as a communication system and used WBE sequence to code digital watermark. Then, using communication detection algorithms, such as sphere decoding or semi-definite relaxation algorithms, to improve digital watermark detection performance. Results of system modeling and simulation show that digital watermark detection based on semi-definite relaxation algorithm can satisfy the requirements of performance and complexity. The improve approach and future development are discussed in the last part of this thesis.